Aa

Sa

b

a

ARR2AA

KAESNVC

1

ipTsipnnisbser

h10

Developmental Cognitive Neuroscience 19 (2016) 223–232

Contents lists available at ScienceDirect

Developmental Cognitive Neuroscience

j o ur nal ho me pa ge: ht tp : / /www.e lsev ier .com/ locate /dcn

uditory evoked potentials to speech and nonspeech stimuli aressociated with verbal skills in preschoolers

oila Kuuluvainena,∗, Alina Leminena,b, Teija Kujalaa

Cognitive Brain Research Unit, Institute of Behavioral Sciences, University of Helsinki, FinlandCenter of Functionally Integrative Neuroscience, Department of Clinical Medicine, Aarhus University, Denmark

r t i c l e i n f o

rticle history:eceived 14 July 2015eceived in revised form3 December 2015ccepted 1 April 2016vailable online 14 April 2016

eywords:uditoryvent-related potential

a b s t r a c t

Children’s obligatory auditory event-related potentials (ERPs) to speech and nonspeech sounds have beenshown to associate with reading performance in children at risk or with dyslexia and their controls. How-ever, very little is known of the cognitive processes these responses reflect. To investigate this question,we recorded ERPs to semisynthetic syllables and their acoustically matched nonspeech counterparts in63 typically developed preschoolers, and assessed their verbal skills with an extensive set of neurocog-nitive tests. P1 and N2 amplitudes were larger for nonspeech than speech stimuli, whereas the oppositewas true for N4. Furthermore, left-lateralized P1s were associated with better phonological and preread-ing skills, and larger P1s to nonspeech than speech stimuli with poorer verbal reasoning performance.Moreover, left-lateralized N2s, and equal-sized N4s to both speech and nonspeech stimuli were asso-

peechonspeecherbal skillshildren

ciated with slower naming. In contrast, children with equal-sized N2 amplitudes at left and right scalplocations, and larger N4s for speech than nonspeech stimuli, performed fastest. We discuss the possibilitythat children’s ERPs reflect not only neural encoding of sounds, but also sound quality processing,memory-trace construction, and lexical access. The results also corroborate previous findings that speechand nonspeech sounds are processed by at least partially distinct neural substrates.

© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

. Introduction

Even before the development of vocabulary, newborns andnfants are biased towards listening to speech over equally com-lex nonspeech sounds (Vouloumanos and Werker, 2004, 2007).his bias lays the foundation to the development of later languagekills via the increasing specialization of the cortex in process-ng speech (for reviews, see Kuhl, 2004; Kuhl et al., 2008). Byreschool age, typically developing children distinguish speech andonspeech sounds effortlessly, and master basic language skillsecessary for learning in a formal school setting. However, little

s known of the underlying processes of speech versus nonspeechound encoding in preschoolers, as no comprehensive studies haveeen conducted in this age group. This study aims to investigatepeech and nonspeech sound processing using cortical auditory

vent-related potentials (ERPs) and their association with neu-ocognitive task performance.

∗ Corresponding author.E-mail address: soila.kuuluvainen@helsinki.fi (S. Kuuluvainen).

ttp://dx.doi.org/10.1016/j.dcn.2016.04.001878-9293/© 2016 The Authors. Published by Elsevier Ltd. This is an open access article

/).

In children under 11 years of age, sounds typically elicit a patternof so-called ‘obligatory’ ERPs labeled P1-N2-N4 or P100-N250-N450 according to their polarity (positive or negative) and latency(100, 250 or 450 ms; e.g. Pihko et al., 2005; Ponton et al., 2000;Shafer et al., 2015). They are identifiable already in neonates toharmonic tones presented at a slow rate, with P1 increasing inamplitude during the first three months, and N2 becoming increas-ingly robust between six and nine months of age (Kushnerenkoet al., 2002). For syllables, P1 is identifiable already at the youngestage group of three-month-olds, whereas N2 emerges at around sixmonths of age, both stabilizing in amplitude and latency by theage of two years (Shafer et al., 2015). P1 amplitude to syllablesincreases again at the age of five, remaining stable after that untilthe age of eight years, whereas N2 amplitude to syllables shows noclear developmental tendencies between ages two and eight years(Shafer et al., 2015).

In contrast to syllables, P1 amplitude to harmonic tones is ofsimilar magnitude at ages four and nine years, decreasing by adult-

hood (Ceponiene et al., 2002) and it decreases steadily for puretones from age seven to adulthood (Bishop et al., 2011; Sussmanet al., 2008) N2 amplitude to harmonic tones decreases betweenages four and nine (Ceponiene et al., 2002) and is stable for pure

under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.

2 Cogni

tNthattp

8see(2ttiobtimtte

v(c(t(i(Nsnrfnn“

sFtHpdiswIbrsa

aapntaN

24 S. Kuuluvainen et al. / Developmental

ones between ages eight and eleven years (Sussman et al., 2008).o results were reported for N4 in these studies. Taken together,

he results suggest that speech and nonspeech sound processingave different developmental trajectories, with turning points atround ages two and five years. However, for a more complete pic-ure, the processing of speech and nonspeech sounds would haveo be compared within the same participants. This has been donereviously with school-aged children, but not with preschoolers.

Previous studies of speech and nonspeech processing in–10-year-old children have given variable results depending ontimulus properties. P1 amplitude was found to be larger for vow-ls than complex or simple tones (Bruder et al., 2011; Ceponienet al., 2001) but smaller for syllables than nonspeech analoguesCeponiene et al., 2005, 2008). The studies of Ceponiene et al. (2005,008) suggest that the child P1 is, in the absence of N1, fusedogether with P2, which in adults is enhanced to discriminationraining (Tremblay et al., 2001) suggesting it reflects neural tun-ng to newly learned sound contrasts. Furthermore, the amplitudef P1 to prototypical vowels was found to correlate inversely withehavioral same-assessment of vowels and with reading speed, sohat children with smaller P1s were more accurate in assessing twodentical vowels “the same” and could also read more words per

inute (Bruder et al., 2011). Therefore, the child P1 was suggestedo reflect both sound detection and speech-nonspeech as well ashe wideness of neural tuning curves to vowel prototypes (Brudert al., 2011).

In the same studies, results for N2 amplitude were similarlyariable. N2 was smaller (Ceponiene et al., 2001) or equal in sizeBruder et al., 2011) for vowels and simple tones when compared toomplex tones, but larger for syllables than nonspeech analoguesCeponiene et al., 2005, 2008). Since the amplitude of N2 elicited byone pips was found to increase with repetition in nine-year-oldsKarhu et al., 1997), larger N2s to complex sounds than vowels werenterpreted as memory-trace build-up for the unfamiliar stimuliCeponiene et al., 2001). In the studies using syllables, N2 and4 behaved similarly, and were suggested to reflect higher-order

ound analysis, such as the content recognition of syllables, scan-ing for access to semantic representations, or short-term memoryetrieval (Ceponiene et al., 2001, 2005, 2008). As N4 was also largeror vowels than simple or complex tones, it is the only compo-ent, which has consistently had larger amplitude for speech thanonspeech sounds, and was thus interpreted as an index of soundspeechness” (Ceponiene et al., 2001, 2005, 2008).

A few studies of preschool children with clinical groups alsotress the usefulness of ERPs as indexes of language development.or example, Lovio et al. (2010) reported diminished P1 peakso syllables in 6-year-old children at risk for dyslexia, whereasämäläinen et al. (2013) reported abnormally large N2s to a shortseudo-word and its nonspeech counterpart in 6-year-old chil-ren who three years later had reading problems. Furthermore,

n a longitudinal study, Espy et al. (2004) presented syllables andinusoidal tones with long, 2.5-4.0 s inter-stimulus intervals (ISI),hich produces the child N1 in addition to the P1-N2-N4 complex.

ncreased N1 amplitudes to both speech and nonspeech stimulietween ages 1 and 4 years were related to poorer pseudo-wordeading at school, whereas decreased N2 amplitudes to nonspeechtimuli between ages 4 and 8 years predicted poorer word readingt school.

Here, our goal was to fill a gap in research by contrasting speechnd nonspeech sound processing in preschoolers, using syllablesnd nonspeech stimuli that were carefully matched for acousticroperties with the speech stimuli. As, to our knowledge, there are

o such previous studies in six-year-olds, our hypotheses are onlyentative. If sound detection quality processing in preschoolers iskin to school-aged children, we will observe smaller P1 but larger2 and N4 responses to syllables than nonspeech sounds (Bruder

tive Neuroscience 19 (2016) 223–232

et al., 2011; Ceponiene et al., 2001, 2005, 2008). We will also analyzethe relationship between cortical responses and neurocognitivetask performance, expecting P1 amplitude to be associated withbetter phonological skills (Bruder et al., 2011), and larger speechthan nonspeech N2/N4s to be associated with better cognitive func-tioning.

2. Methods

2.1. Participants

Originally, 94 typically developed monolingual Finnish-speaking children participated in a longitudinal study of preschoollanguage abilities and later reading performance. The currentstudy consists of the preschool data of 63 children (33 boys; 3left-handed, 1 ambidextrous) that remained after the exclusion ofthe data of 31 children due to cancellation of participation (N = 12),a PIQ lower than the set limit of 85 (N = 1), later discovery ofunclear family history of neurological problems (N = 1), excessivealpha band activity (N = 11) or motor artifacts (N = 8). The mean ageof the children was 6 years 6 months (range 6 years 0 months–7years 0 months), and they had an average of 80 (range 8–156) daysof preschool teaching prior to the EEG experiment. All childrenwere born full-term and had reportedly normal hearing. Mostparents of the children had completed high school (fathers 73%,mothers 86%), and had college or university education (fathers59%, mothers 71%), and were employed (fathers 90%, mothers76%). The family background of these children is typical to theFinnish metropolitan capital area (Official Statistics of Finland(OSF), 2013).

The study was approved by the Ethical Board of Helsinki andUusimaa Hospital District. Written consent was obtained from par-ents and oral consent from the child.

2.2. Stimuli and paradigm

The semi-synthetic CV syllables and their acoustically matchednonspeech counterparts used as stimuli were created using theSemisynthetic Speech Generation (Alku et al., 1999) method. Vow-els/i/and/e/were compiled by extracting a glottal excitation froma natural speech signal, obtaining the desired formant structurewith a digital all-pole filter, and adding the filtering effect of thevocal tract to the model. The F0 was 101 Hz for both vowels. For/i/,the lowest four formant frequencies were 410, 2045, 2260, and3320 Hz and for/e/, 320, 2240, 2690, and 3275 Hz. The unvoicedplosives/k/and/p/were extracted from syllable/ke:/and the shortword/pito/, and inserted to the beginning of the semi-syntheticvowels to create standard stimulus syllables/pi/and/ke/. The totalduration of the standard stimulus was 170 ms (12 ms consonantand 158 ms vowel sections, including 5 ms rise and fall times), andits intensity set to approximately 55 dB SPL.

The nonspeech sounds were created by mimicking the glot-tal flow of the semi-synthetic syllables with a carefully controlledimpulse train, so that the F0 was equal with the speech stimuli.Linear predictive coding (LPC; Rabiner and Schafer, 1978) of a pre-diction order of 10 was used to match the spectral envelope tothat of the speech sound. Then the impulse train was used as anexcitation to an all-pole filter which modeled only the second for-mant, i.e., the all-pole vocal tract consisted of a single resonance at2240 Hz and 2045 Hz for the nonspeech counterpart of the/i/and/e/,respectively. LPC coding of a prediction order of 50 was used to

model/p/and/k/, exciting it with random noise. The nonspeech syl-lables were formed by combining the corresponding nonspeechconsonant and vowel counterparts (see Fig. 1). The speech and non-speech stimuli were thus matched in terms of duration, F0, intensity

S. Kuuluvainen et al. / Developmental Cognitive Neuroscience 19 (2016) 223–232 225

F CV syo

aah

pvfFi22b

prd5Sbti

ig. 1. Sound waveforms and spectrograms of the used stimuli. The semisyntheticn the lower half of the figure.

nd spectral envelope behavior, with only one of them recognizeds speech, and the other described as “the sound of a buzzer or aorn” (Kuuluvainen et al., 2014).

The stimuli were presented in a multi-feature MMNaradigm with five deviant stimuli (consonant/counterpart,owel/counterpart, sound F0, duration, and intensity) derivedrom the standard sounds, all syllables having no meaning ininnish. The results for deviant stimulus responses are reportedn another study (Kuuluvainen et al., 2015). In addition, 46 novel00 ms complex sounds (telephone ringing, etc.; see Sorokin et al.,010 for further details) were included to study attention, resultseing reported elsewhere.

The standard sounds/pi:/,/ke:/, nonspeech ‘pi’, and ‘ke’ wereresented in separate, pseudo-randomised blocks with theirespective deviants, resulting in a total of 538 stimuli (10 stan-ards excluded from the analysis, followed by 528 stimuli, of which0% were standards, 8.33% of each deviant, 8.33% of novel sounds).tandards after novel sounds were excluded from the analysis. Each

lock was presented twice, resulting in eight four-minute blocks inotal. The order of the blocks was counterbalanced between partic-pants.

llables/pi:/and/ke:/are on the upper, and their respective nonspeech counterparts

2.3. EEG data recording and experiment procedure

EEG was recorded with 28 active electrodes (BrainVisionQuickAmp & ActiCap, Brain Products, Germany), with four activeelectrodes placed on left and right mastoids, near the outer can-thus of the right eye (horizontal electro-oculogram, HEOG) and thenose. Two passive bipolar electrodes were used to record the verti-cal electro-oculogram (VEOG). Recording reference was the averagesignal of all electrodes, sampling rate 250 Hz, and frequency bandfrom DC to 50 Hz.

During the recording, the children sat in a comfortable chairin an electrically shielded and sound-attenuated room. Stimuliwere delivered to both ears via Sony Professional MDR-7506headphones, while they watched a self-selected, silent video. Thechildren were instructed to watch the program carefully for ques-tions of its content after each block, and not to pay attentionto the stimuli. The overall quality of the children’s reports wasgood, indicating they had attended to the movie. After every two

blocks, the child could choose a small item (toys, pens, stickers)to keep up motivation. During a longer mid-experiment break, thechild was given juice and cookies, and after finishing, could choose

226 S. Kuuluvainen et al. / Developmental Cogni

Table 1Children’s neurocognitive test results.

Mean (SD) Min-Max

Age (years) 6.5 (0.2) 6.0–7.0Total testing time (min) 138 (33) 100–360

Reasoning skillsPIQ 103 (13) 85–139VIQ 107 (10) 81–128

General language skillsBNT (correct) 39 (5.0) 25–48Comprehension of instructions (SPa) 10.6 (2.6) 4–17

Working memorySentences (SP) 11.3 (2.2) 5–15Numbers (SP) 10.8 (2.3) 5–15

Phonological skillsPseudowords (SP) 9.2 (1.8) 4–13Phonological processing (SP) 10.0 (2.3) 6–16

Rapid Alternating NamingColours, time (s) 63 (14) 43–113Numbers, time (s) 64 (19) 33–141Letters, time (s) 60 (17) 31–98Objects, time (s) 75 (14) 42–114Numbers & Letters, time (s) 57 (16) 36–89Colours, errors 1.9 (1.9) 0–8Numbers, errors 3.1 (4.3) 0–27Letters, errors 3.8 (4.1) 0–18Objects, errors 3.1 (2.9) 0–13Numbers & Letters, errors 4.1 (5.3) 0–28

Reading skillsb

Letter naming (correct) 24 (6.2) 3–29Syllable reading (correct) 7.5 (8.1) 0–18Word reading (correct) 12.8 (18.2) 0–74

SD = Standard deviation, SP = Standard points, BNT = Boston naming test.a SPs are based on age norms, where the normative age mean is 10 and standard

d

r

bm

2

1vo(8W2trw

eeAKtpioo(Itc

eviation 3 points.b The maximum score for letter naming is 29, for syllable reading 18 and for word

eading 90.

etween movie tickets and a monetary compensation. The experi-ent, including preparation and breaks, lasted approximately 2 h.

.4. Neurocognitive tests

The neurocognitive testing session was carried out on average8 days (range 1–64) after the EEG. We obtained performance anderbal intelligence quotients (PIQ and VIQ) with the Finnish versionf the Wechsler Preschool and Primary Scale of Intelligence—IIIWPPSI-III; (Wecshler, 2009). If the child’s preschool PIQ was under5 or VIQ under 75, they were reassessed a year later, using theechsler Intelligence Scale for Children—IV (WISC-IV; Wechsler,

010) This was done to minimize shyness or lack of motivation ashe reason for a below-threshold PIQ/VIQ score in preschool. Aftereassessment, all but one child (data discarded) had PIQ and VIQithin limits (see Table 1).

In addition to reasoning abilities, we assessed the children’s gen-ral language abilities with the Boston Naming Test (BNT; Kaplant al., 1997), and the Comprehension of instructions subtest from

Developmental Neuropsychological Assesment—II (NEPSY-II;orkman et al., 2008) which tests the child’s ability to point pic-

ures following auditory instructions. In the BNT, the child namesictures, and is given a cue, e.g. the common use of the object, only

f s(he) clearly misinterprets the picture. The score is the numberf correctly named pictures. Verbal short term and working mem-ry were assessed with the The Digit span subtest from WISC—III

Wechsler, 1999) and the Sentence repetition subtest from NEPSY-I (Korkman et al., 2008) In the Digit span subtest the child is askedo repeat numbers first in the same, and then in reverse order. Eachorrectly repeated sequence scores one raw point. In the Sentence

tive Neuroscience 19 (2016) 223–232

repetition subtest the child hears more and more complex sen-tences and is to repeat them word-for-word. Correctly repeatedsentences score two raw points, and those with one or two mistakesone point.

Phonological skills were assessed using the Phonological process-ing subtest from NEPSY-II and the Repetition of nonsense wordssubtest from NEPSY I (Korkman et al., 1997). In the Phonologicalprocessing subtest the child hears three words with matching pic-tures and is asked to point the picture for a word containing theseparately given target syllable. Later in the test the child is askedto first repeat a word and then either omit or change one or morephonemes to form a new word. Each correct answer scores onepoint. In the Repetition of nonsense words subtest the child hearsnonwords, and is asked to repeat them as accurately as possibly,each correctly repeated nonword scoring one point. The children’snaming speed was tested with the Rapid Automatized Naming test(RAN; Ahonen et al., 2003), in which the child is asked to name itemsas fast and accurately as (s)he can. We used the first five tables(colours; numbers; letters; objects; numbers & letters; colours,numbers & letters) which contain five rows of ten items each. Thechild’s time as well as errors made during naming are recorded.Finally, we assessed prereading skills with letter naming (the Finnishalphabet in random order) and syllable reading (18 non-word syl-lables au, is sa, eu, ki, tu, oi, pe, as, iu, ke, pi, isu, upe, ako, eso, uti,oke, given to the child one per time in a small card to be read outloud). Children mastering at least half of the syllables proceededto the Finnish LukiLasse (Häyrinen et al., 1999) word reading task(measuring correctly read words in 120 s). For standardized tests(WPPSI-III, WISC-IV, Nepsy I and II) raw scores were converted toage-appropriate standardized scores (mean score in the normativeage group being 10, and standard deviation 3 points).

We used two different testing orders, balancing auditory-onlyand visually guided subtests. Orders were assigned balancing forgender, and no differences emerged between scores from differentorders (t (58–61) = −0.05-1.8; p > 0.05). The neurocognitive test-ing session lasted 2.5 h on average, including three breaks, withsimilar rewards as the EEG session. Parents wishing so receivedfeedback of the age-typicality of their child’s performance over thephone, emphasizing that single-time test results of small childrencan be unreliable due to attention, motivation and fatigue effects,and that children’s cognitive skills often develop in different paces.Parents could discuss the results also with one of the authors (S.K.,a licensed psychologist). If the parents wished to have the childassessed further, they were guided to contact the Finnish healthcare system.

For all investigated skills where age-appropriate norms wereavailable, the group average was close to the age-typical standard-ized mean score of 10 points (100 points for VIQ and PIQ; seeTable 1). For the RAN and BNT, conversion to standard points wasnot possible, but the average naming times and numbers of errorswere comparable to those reported in the normative data for six-year-olds (Ahonen et al., 2003; Kaplan et al., 1997). No individualscore was set as exclusion limit apart from those of VIQ and PIQ, andthe range of performance varied therefore from poor to excellentin different tasks.

2.5. EEG data analysis

BESA 6.0 (Besa GmbH, Germany) software was used to prepro-cess the raw data individually for each block. After interpolatingmissing or bad channels, blink artifacts were removed using the

BESA Principal Component Analysis (PCA) Tool, the thresholds set at75 �V for both VEOG and HEOG channels. If blink patterns were stillobserved, the “Define Artifact Topography” Tool was used, the blinkdefinition criteria being a PC with bipolar topography across the

S. Kuuluvainen et al. / Developmental Cognitive Neuroscience 19 (2016) 223–232 227

Table 2Test scores used in rANCOVA.

Mean (SD) MIN 25th 50th 75th MAX

Phonological skills (%)a 58.5 (8.2) 41.1 52.4 58.0 63.3 75.2Prereading skills (%)a 62.5 (30.1) 5.2 38.0 48.3 95.5 100Rapid Alternating Naming (s)b 69.2 (12.3) 44.5 62 70 77 102Verbal Short Term Memory (%)a 53.3 (6.6) 34.5 49.5 53.9 60.0 66.4PIQ 103 (13) 85 91 102 111 139VIQ 107 (10) 81 100 109 115 128

SD = Standard deviation; MIN = minimum; 25th, 50th and 75th percentiles;M

hi

mfidttMasumftc

2

Czcse

ttaarssb(tw

oaNra(nndetEnd

AX = maximum.a Average percent correct across two subtests.b Average naming speed in seconds across two subtests.

orizontal axis of the eyes, explaining at least 85% of the variancen the selected data.

The data were then re-referenced to mastoids (see Supple-ent A for nose-referenced data) and epoched (−100–500 ms),

ltered (1–30 Hz), and corrected for baseline (−100–0 ms). Theata were averaged rejecting epochs with (a) peak-to-peak ampli-ude (s) greater than 75 �V. These blockwise averaged ERPs werehen exported as ASCII multiplexed files to EEGLAB (Delorme and

akeig, 2004), where averages over blocks were formed. The aver-ge number of accepted trials was 667 (range 437–786) for thepeech and 663 (range 482–767) for the nonspeech standard stim-li. P1 was defined as the positive, and N2 and N4 as the negativeaxima, at 50–150 ms, 150–300 ms, and 300–450 ms, respectively,

rom the Cz electrode. Amplitudes were quantified by calculatinghe individual mean response amplitude in a ±5 ms time windowentered at the peak of the group grand-mean response.

.6. Statistical analyses

The statistical significance of the ERPs was determined at thez electrode, using a t-test comparing the amplitude values toero. For each of the three speech–nonspeech response pairs, weonducted a three-way rANOVA of condition (speech versus non-peech), anterior-posterior, and lateral distribution, using ninelectrodes (F3, Fz, F4, C3, Cz, C4, P3, Pz, and P4).

Averaged scores across tests and electrodes were used to inves-igate the relationship between ERPs and neurocognitive tests. Forhe ERPs, the mean amplitudes of three electrodes in left (F3, C3, P3)nd three in right (F4, C4, P4) scalp locations were calculated. PIQnd VIQ scores were used unaltered. We used percentage correctaw scores from phonological processing and repetition of non-ense word for phonological skills, from digit span forward andentence repetition for vSTM, and from letter naming and sylla-le reading for prereading skills. For the RAN score, naming speedsin s) of colours and objects were averaged, since reliable data forhese two was available also for those children having problemsith numbers or letters (see Table 2 for details of all scores).

For each of the six scores, a two-way repeated measures analysisf covariance (rANCOVA) of condition (speech versus nonspeech)nd laterality (left versus right) was conducted for P1, N2, and4 separately, using the test score as covariate. To interpret the

esults, Pearson correlation coefficients were calculated betweenll test scores and between each test score and: (1) ERP measure;2) amplitude differences between left and right, and speech andonspeech conditions, always subtracting the amplitude of right oronspeech from left or speech ERP amplitude, respectively; (3) theifference of the left and right speech-nonspeech amplitude differ-nce. This last measure was at its greatest when the amplitude of

he left speech ERP was larger than the corresponding nonspeechRP, with an opposite pattern on the right. The amplitudes of theegative ERPs (N2 and N4) were reversed to keep the correlationirections constant across all three ERPs.

Fig. 2. ERPs to standard stimuli. Grey areas highlight the time windows of amplitudequantification for each response. Respective topographic maps are presented belowthe waveforms.

To correct for observed violations of sphericity, we appliedGreenhouse-Geisser corrections (original degrees of freedomreported with � and corrected significance). Statistically significantinteractions were further examined with Bonferroni-corrected posthoc tests. All analyses were conducted with SPSS Statistics 22 (IBM,U.S.A.).

3. Results

3.1. ERP amplitudes

All standard stimulus responses significantly differed from zero(t(62) = −15.5–19.9, p < 0.001); see Table 3). There was a main effectof condition for all three responses (P1: F(1,62) = 13.6, p < 0.001,partial ETA2 = 0.18, observed power Po=0.95; N2: F(1,62) = 9.1,p = 0.004, partial ETA2 = 0.13, Po = 0.84; N4: F(1,62) = 7.8, p = 0.007,partial ETA2 = 0.11, Po = 0.79). For P1 and N2 this resulted fromlarger responses to nonspeech than speech stimuli, whereas for N4the effect was the opposite (see Fig. 2). Further, there was a condi-tion × anterior-posterior × laterality interaction for the P1 response(F(4,248) = 2.9, � = 0.78, p = 0.036, partial ETA2 = 0.044, Po = 0.70),which resulted from larger responses for nonspeech than speechstimuli in the middle and right than left and anterior electrodesites. Also, there was a statistically significant condition × anterior-posterior interaction for the N2 response (F(2,124) = 11.6, � = 0.70,p < 0.001, partial ETA2 = 0.16, Po = 0.99), which resulted from fronto-centrally larger responses for nonspeech than speech stimuli (seeFig. 2).

3.2. Covariance analysis for test scores and ERPs

Better performance in phonological and prereading skillswere associated with larger P1 amplitudes at the left thanthe right scalp locations, especially for the speech stim-uli (condition × laterality × phonological skills interaction:

228 S. Kuuluvainen et al. / Developmental Cognitive Neuroscience 19 (2016) 223–232

Table 3ERP latencies, amplitudes and statistical significance.

Speech Nonspeech

Latency (Cz; ms) Amplitudea (Cz; �V) t(62) Sign.b (p) Latency (Cz; ms) Amplitude (Cz; �V) t(62) Sign. (p)P1 108 6.0 (2.4) 19.9 <0.000 108 6.6 (2.7) 19.6 <0.000N2 216 −2.5 (1.8) −11.4 <0.000 216 −3.0 (1.7) −13.4 <0.000N4 308 −3.3 (1.7) −15.5 <0.000 304 −2.9 (1.9) −12.2 <0.000

a Mean, standard deviation in brackets.b Statistical significance of the response compared to zero.

N2 and naming speed

-1.5

0

1

.5

-1

-.5

0.51

-.5-1

N4 and naming Speed

NonspeechSpeech

LeRight

L-R differenceL-R difference*

S-X difference S-X difference †

P1 and phonological skills P1 and prereading skills

0

1

.5

-.5

0.51

-.5-1

LeRight

NonspeechSpeech

L-R difference*

S-X difference S-X difference S-X difference †

L-R difference

P1 and verbal reasoning

G1 G4G3G2G1 G4G3G2 G1 G4G3G2

G1 G4G3G2G1 G4G3G2

L-R difference*

ɥV

ɥV

group group

Fig. 3. The relationship between test performance and cortical responses. For illustrative purposes, children were divided into quartile groups (G1-G4) according to testperformance, so that children in G1 were the lowest 25% and children in G4 the highest 25% of scorers for the skill in question (see Table 2 for details; note that groupsare reversed for naming speed, as good performers are faster). The upper images show amplitude difference between left and right scalp locations, positive values meanl ces bes negatS

Fdpa

arger responses on the left than right. The lower images show amplitude differenpeech than nonspeech stimuli. Differences were calculated for reversed values for

= speech, X = nonspeech.

(1,61) = 6.2, p = 0.016, partial ETA2 = 0.092, Po = 0.69, and con-

ition × laterality × prereading skills interaction: F(1.61) = 5.8,

= 0.019, partial ETA2 = 0.087, Po = 0.66). Furthermore, there was marginally significant condition × VIQ interaction (F(1,61) = 3.0,

tween speech and nonspeech responses, positive values mean larger responses toive responses. Error bars represent ± 1 standard error of the mean. L = left, R = right,

p = 0.086, partial ETA2 = 0.048, Po = 0.40), which resulted from an

increased difference in nonspeech versus speech P1 amplitudesassociating with poorer VIQ scores (see Fig. 3). Furthermore, slowerperformance in RAN was associated with larger N2 amplitudes in

Cogni

tttaPsss

sitTwpbvlcfrl

4

nusatsniiaso

4s

fsbstseciePPm2cnswe

as

S. Kuuluvainen et al. / Developmental

he left than right scalp locations (laterality × RAN speed interac-ion: F(1,61) = 7.0, p = 0.010, partial ETA2 = 0.10, Po = 0.72). Finally,here was a marginally significant overall effect of RAN speednd N4 amplitude (F(1,61) = 3.8, p = 0.057, partial ETA2 = 0.058,o = 0.48), faster naming performance being associated with largerpeech than nonspeech N4 amplitudes (Fig. 3). There were notatistically significant results for the rANCOVAs with PIQ or verbalhort term memory as a covariate.

From the neurocognitive skills associated with ERP measures,ignificant intercorrelations were found between verbal reason-ng, phonological, and prereading skills, whereas none of thesehree correlated with RAN speed (r = 0.340–0.640, see Table 4).he correlations between neurocognitive skills and ERP measuresere in line with the results of the rANCOVA: Phonological andrereading skills correlated with P1 left-speech right-nonspeechias (r = 0.304 and 0.205, p = 0.016 and 0.019, respectively), anderbal reasoning was marginally significantly correlated with P1eft speech-nonspeech difference (r = 0.248, p = 0.050). RAN speedorrelated significantly with nonspeech N2 left-right amplitude dif-erence (r = 0.375, p = 0.002) and inversely with the amplitude ofight speech N4 (r = −0.290, p = 0.021) as well as marginally witheft speech N4 (r = −0.229, p = 0.071; see Table 5).

. Discussion

We investigated cortical pre-attentive processing of speech andonspeech sounds in typically developed six-year-old children,sing semi-synthetic syllables and their acoustically matched non-peech counterparts. Our results suggest that already at preschoolge, the neural substrates responsible for sound detection and con-ent analysis are distinct for speech and equally complex nonspeechounds. We found that P1 and N2 were smaller for speech thanonspeech sounds, whereas the N4 showed an opposite pattern,

ndicating that sound “speechness” quality, and/or sound famil-arity, is detected and processed at very early stages of auditorynalysis. Furthermore, individual differences between responses topeech and nonspeech stimuli were associated with several aspectsf verbal cognitive functioning.

.1. The relationship of P1, phonological and verbal reasoningkills

The “speechness” quality of a sound, or alternatively soundamiliarity, was cortically detected already after the first 100 ms ofound processing, as reflected by the smaller P1 amplitude to sylla-les than nonspeech counterparts. This is consistent with previoustudies in school-aged children, showing smaller P1s to syllableshan complex harmonic sounds, which was explained by the non-peech stimuli being perceptually more salient and thus moreasily detectable (Ceponiene et al., 2005, 2008). However, since wearefully matched the speech and nonspeech stimuli, differencesn stimulus complexity or acoustic properties, including stimulusnergy, cannot explain the observed differences. However, the child1 has been suggested to be merged with the equivalent of adult2 (Ceponiene et al., 2005, 2008), which in turn has been linked toemory trace formation for unfamiliar contrasts (Tremblay et al.,

001). It is thus possible that the larger P1 responses to nonspeechounterparts than speech sounds reflect the build-up or mainte-ance of a memory trace for these equally complex but unfamiliarounds. This is also supported by the findings of Bruder et al. (2011),ho showed that more accurate vowel prototype memory traces,

stablished prior to the experiment, associated with smaller P1s.Differences between P1s to speech and nonspeech sounds

ppeared at the right and midline, but not at left and anterior sites,uggesting a right-hemisphere bias in the processing of nonspeech

tive Neuroscience 19 (2016) 223–232 229

compared to speech sounds. This is consistent with previous studiesof speech and nonspeech processing in adults, indicating a right-ward bias for nonspeech processing (for reviews, see McGettiganand Scott, 2012; Tervaniemi and Hugdahl, 2003). However, P1 later-alization effects were not observed in the previous studies of speechand nonspeech processing in school-aged children (Bruder et al.,2011; Ceponiene et al., 2001, 2005, 2008). One possible explana-tion is that the smaller group sizes of these studies (N = 11–14) didnot allow for lateralization effects to emerge.

In the current study, larger P1s to speech sounds on the leftthan right, and larger P1s to nonspeech sounds in the right than leftscalp locations, were related to better phonological and prereadingskills. The children performing poorest in these tasks showed theopposite pattern, with larger nonspeech P1s at left than right scalplocations. Note, that the similarity of the associations of phono-logical and prereading skills with P1 is probably explained by therelatively strong inter-correlation of these measures. Furthermore,larger P1 differences between nonspeech and speech sounds, espe-cially left to midline, were associated with poorer verbal reasoningskills. In contrast, children performing best had equal-sized P1s toboth stimuli. Hence, we propose that P1 reflects processes relatedto successful establishment of long-term phonological representa-tions especially in the left hemisphere, which in turn is associatedwith better phonological skills and with more mature verbal func-tioning. In contrast, if the processes behind P1 elicitation are biasedtowards the right hemisphere for speech, or are engaged in exten-sive processing of nonspeech sounds in the left hemisphere, verbaland especially phonological skills fall below age expectations.

The physiological explanation behind this might be in neuronalover-representation of native language speech sounds, resultingin erroneous assignment of relevance. This would lead to neuralengagement in the left cortical areas to the nonspeech counter-parts, which have a speech-like acoustic structure but are novelto the children. This interpretation is also consistent with studiesshowing smaller P1s to associate with increased maturation of thecortex. P1 amplitude elicited to pure tones showed steady decreasefrom age seven onwards (Bishop et al., 2011; Sussman et al., 2008),and smaller P1s to vowel prototypes reflected more precision inneural tuning curves (Bruder et al., 2011). These possibilities couldbe further investigated by using also non-native speech sounds asin Bruder et al. (2011), or with a learning paradigm, in which chil-dren would be taught to recognize and discriminate the nonspeechsounds as communicative.

4.2. The relationship of N2 and N4 with speed of lexical access

Unlike 8–10 year-old children (Ceponiene et al., 2005, 2008),the preschoolers of the current study had larger N2 responsesto the nonspeech analogues than to the CV syllables. Our resultis, however, similar to that obtained by Ceponiene et al. (2001),who reported larger N2s to complex tones than vowels inschoolchildren. They then speculated that N2 reflects memory-trace formation, since Karhu et al. (1997), had previously shownthat N2 amplitude increases with stimulus repetition. In this casethe larger nonspeech than speech N2 amplitudes in our studywould be explained by a process similar to that explaining the P1results.

A second possible explanation to the different results betweenpreschoolers and schoolchildren is the maturation of speech- andnonspeech-elicited N2. Earlier longitudinal studies suggest that N2elicited by vowels is relatively stable in amplitude between ages

five and eight years (Shafer et al., 2015), whereas N2 elicited byharmonic tones decreases in amplitude from age four to nine years(Ceponiene et al., 2002). It is thus possible that during the transi-tion to school-age, the cognitive processes underlying N2 mature,

230 S. Kuuluvainen et al. / Developmental Cognitive Neuroscience 19 (2016) 223–232

Table 4Intercorrelations of neurocognitive measures.

Neurocognitive measure PIQ VIQ PHON PRER VSTM RAN

Performance reasoning (PIQ) 1Verbal reasoning (VIQ) .421b 1Phonological skills (PHON) 0.011 .340b 1Prereading skills (PRER) 0.10 .349b .640b 1Verbal short term memory (VSTM) .272a .447b .426b .342b 1Naming speed (RAN) 0.073 −0.15 −0.072 −0.21 −0.19 1

a Correlation significant at p < 0.05 level (2-tailed).b Correlation significant at p < 0.01 level (2-tailed).

Table 5Correlations between neurocognitive skills and ERP measures.

Neurocognitive skill ERP measure ra Sig.b

Verbal reasoning P1 Left Speech-Nonspeech Differencec 0.248 0.050Phonological processing P1 Left-Speech Right-Nonspeech Biasd 0.304 0.016Prereading skills P1 Left-Speech Right-Nonspeech Bias 0.295 0.019Naming speed N2 Nonspeech Left-Right Differencee 0.375 0.002Naming speed N4 Speech Leftf −0.229 0.071Naming speed N4 Speech Right −0.290 0.021

Note that the amplitudes of N2 and N4 were reversed to keep correlation directions consistent.a Pearson’s correlation coefficient; note that negative ERPs were reversed for analysis.b Two-tailed significance.c Amplitude difference of left speech and nonspeech P1s.d Difference of left and right amplitude differences of speech and nonspeech P1s (maximal when speech P1 is largest at left and smallest at right, and nonspeech P1 vice

v

ca

atw6btst2mioCa−Noicpb

eiaTstpatvtr

ersa).e Difference of left and right nonspeech N2s.f The amplitude of left speech N4.

hanging the response pattern to favor syllables over nonspeechnalogues.

A third explanation of the different N2 results in our studynd the previous studies of Ceponiene et al. (2005, 2008) arehe stimuli used. Compared to Ceponiene et al. (2005, 2008),ho used CV syllables and nonspeech counterparts with only

0–80 ms vowel/counterpart sections, the vowels of our CV sylla-les/counterparts were relatively long (158 ms). It is possible thathe longer vowel length of 158 ms in the current study led to smallerpeech than nonspeech N2s. This interpretation is supported byhe results of Ceponiene et al. (2001), who reported larger N2s to60 ms long complex tones than vowels in schoolchildren. Further-ore, the studies of Shafer et al. (2005) and Datta et al. (2010)

n 8–10 year-olds used short (50 ms) and long (250 ms) versionsf the same vowels, and shared six of their nine control children.omparing across these two studies, N2 to the short vowel/i/waspproximately twice the size as to the long version (−4 versus2 �V), whereas P1 sizes were more or less equal (−5 �V) and4 was elicited only to the long but not the short vowel. As, tour knowledge, there are no longitudinal within-participant stud-es of speech versus acoustically matched nonspeech processingovering both preschool and school-aged children, nor studies com-aring vowels of different lengths directly, these questions have toe addressed by further research.

Interestingly, larger N2s in the left than the right scalp areas,specially for nonspeech sounds, were associated with slower nam-ng speed, which is a strong predictor of later reading ability (for

review, see Kirby et al., 2010; see also Dandache et al., 2014;orppa et al., 2010). This effect emerged for both speech and non-peech sounds, suggesting that it reflects a domain-general ratherhan speech-specific neural process. The result is consistent withrevious research by Hämäläinen et al. (2013) reporting larger N2mplitudes in preschool children at risk for dyslexia than con-

rols using short (300 ms), intermediate (360 ms) and long (460 ms)ersions of the disyllabic pseudoword/a:ta/, varying the length ofhe gap between the two syllables. ERPs were also recorded toespective nonspeech sounds, where the/a/vowels were replaced

with harmonic tones comprised of seven sinusoidals, consonantportion staying the same. The difference between at-risk and con-trol children appeared at left scalp locations for the short 300 msspeech stimulus and for all nonspeech stimuli, and at the right scalplocations only for the intermediate-length nonspeech stimulus. InHämäläinen et al. (2013), larger speech and nonspeech N2 ampli-tudes at preschool were associated with slower and less accuratereading at school in both groups, similarly to those reported by Espyet al. (2004). The results of our study, showing larger N2s to asso-ciate with slower naming speed, suggest that these links betweenpreschool N2 amplitude and later reading performance might bemediated via lexical access. They imply that smaller N2s reflect amore mature cognitive process, or alternatively, larger responsesreflect a maturational delay.

Similarly as in the school-aged children of Ceponiene et al.(2001, 2005, 2008), preschooler’s N4 amplitudes in our study werelarger for speech than nonspeech sounds. Furthermore, larger N4sto speech than nonspeech sounds were associated with faster RANperformance, the amplitudes of N4s to speech sounds correlatinginversely with naming speed, so that children with shorter namingtimes had larger N4s to speech sounds. These results are consistentwith the earlier interpretations of the processes behind the childN4 response. It seems to be associated with further processing ofsounds identified as speech, and is likely to reflect automatic scanfor access or success of retrieval of lexical representations linked tothe perceived sounds.

4.3. Study limitations and future directions

The current study holds a few limitations. First, the directionof attention was controlled only by giving the children a task ofattending to the videos and answering questions, and there was nocontrol condition in which the children would have been asked to

attend to the stimuli. Consequently, the effects of possible covertattention to the stimulus stream cannot be excluded as an expla-nation of the results. Shafer et al. (2007) studied 8–10-year-oldchildren with SLI and their controls, and found that directing atten-

Cogni

titsdacsbwtttu

tubgTfictercstws

5

ctPstcnsaocobrid

C

A

PMwwaEt

S. Kuuluvainen et al. / Developmental

ion to speech stimuli reduced P1 amplitude, presumably via thenfluence of elicited processing negativity Nd. Furthermore, whenhe children were watching a silent video and asked not to attend totimuli, P1 amplitudes were smaller for typically developing chil-ren than those with SLI, suggesting the former still allocated somettention to the speech stimuli even when requested not to. In theurrent study, smaller P1s to speech than nonspeech stimuli wereeen at the children having lowest verbal reasoning scores, whereasest performers had equal-sized P1s to both conditions. If the resultas to be interpreted as in Shafer et al. (2007), it would imply that

he verbally less mature children would allocate more attentiono speech than their more mature peers, which is in contrast tohe original finding. The role of attention, nevertheless, remainsnresolved, and should be controlled in future studies.

Secondly, the stimuli included only semi-synthetic speech andheir nonspeech counterparts, based on Finnish language. Thenderlying functions should thus be investigated further usingoth natural and semi-synthetic speech sounds from different lan-uages, and expanding to other auditory domains such as music.his would further illuminate the speech and language speci-city of the current findings. Furthermore, learning paradigmsould reveal if memory trace formation is indeed reflected inhe obligatory responses. Longitudinal studies looking at differ-nt age groups could show if the functional significance of theesponses changes in typical development, and the inclusion oflinical groups would show if the associations also hold for atypicalpeech processing. Finally, methods allowing for source localiza-ion, such as magnetoencephalography (MEG) or high-density EEGith anatomical images, could elucidate more precisely the neural

ubstrates involved in the generation of the ERPs.

. Conclusion

Based on the current results, we propose that syllables andorresponding nonspeech sounds are processed by at least par-ially different neural substrates in preschoolers’ auditory cortices.1 and N2 responses were enhanced for nonspeech compared topeech sounds, suggesting that the early stages of neural detec-ion at this age are more responsive to unfamiliar than familiaromplex auditory material. Enhancement of speech compared toonspeech sounds was observed only for N4. The investigated ERPseem to reflect different cognitive processes: the P1 appears to bessociated with phonological and prereading skills as well as withverall verbal functioning, whereas N2 and N4s seem to be asso-iated with the speed of access to lexical representations. Overall,ur results suggest that speech and nonspeech-elicited ERPs cane useful measures of children’s cortical functioning. Since theseesponses can be recorded without any active role of the partic-pant, they could illuminate speech vs. nonspeech functions andevelopment even in smaller children.

onflict of interest

The authors declare no conflicts of interest.

cknowledgments

We thank the children and families for participating, andirita Uuskoski, Piia Turunen, Lilli Kimppa, Saila Seppänen, Hennaarkkanen, and Roope Heikkilä for their help in data collection, asell as Prof. Paavo Alku for his comments on the stimuli. This work

as supported by the Academy of Finland (grants number 128840

nd 1276414), the Finnish Cultural Foundation, Ella and Georghrnrooth Foundation, Emil Aaltonen Foundation, Kone Founda-ion, Lundbeck Foundation, and Jane and Aatos Erkko Foundation.

tive Neuroscience 19 (2016) 223–232 231

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.dcn.2016.04.001.

References

Ahonen, T., Tuovinen, S., Leppäsaari, T., 2003. Nopean Sarjallisen NimeämisenTesti. Niilo Mäki Instituutti & Haukkarannan koulu, Jyväskylä (RapidAlternating Naming Test).

Alku, P., Tiitinen, H., Näätänen, R., 1999. A method for generating natural-soundingspeech stimuli for cognitive brain research. Clin. Neurophysiol. 110 (8),1329–1333.

Bishop, D.V.M., Anderson, M., Reid, C., Fox, A.M., 2011. Auditory developmentbetween 7 and 11 years: an event-related potential (ERP) study. PLoS One 6(5), e18993, http://dx.doi.org/10.1371/journal.pone.0018993.

Bruder, J., Leppänen, P.H.T., Bartling, J., Csépe, V., Démonet, J.-F., Schulte-Körne, G.,2011. An investigation of prototypical and atypical within-category vowelsand non-speech analogues on cortical auditory evoked related potentials(AERPs) in 9 year old children. Int. J. Psychophysiol. 79 (2), 106–117.

Ceponiene, R., Shestakova, A., Balan, P., Alku, P., Yiaguchi, K., Naatanen, R., 2001.Children’s auditory event-related potentials index sound complexity andspeechness. Int. J. Neurosci. 109 (3), 245.

Ceponiene, R., Rinne, T., Näätänen, R., 2002. Maturation of cortical soundprocessing as indexed by event-related potentials. Clin. Neurophysiol. 113 (6),870–882.

Ceponiene, R., Alku, P., Westerfield, M., Torki, M., Townsend, J., 2005. ERPsdifferentiate syllable and nonphonetic sound processing in children and adults.Psychophysiology 42 (4), 391–406.

Ceponiene, R., Torki, M., Alku, P., Koyama, A., Townsend, J., 2008. Event-relatedpotentials reflect spectral differences in speech and non-speech stimuli inchildren and adults. Clin. Neurophysiol. 119 (7), 1560–1577.

Dandache, S., Wouters, J., Ghesquière, P., 2014. Development of reading andphonological skills of children at family risk for dyslexia: a longitudinalanalysis from kindergarten to sixth grade. Dyslexia 20 (4), 305–329, http://dx.doi.org/10.1002/dys.1482.

Datta, H., Shafer, V.L., Morr, M.L., Kurtzberg, D., Schwartz, R.G., 2010.Electrophysiological indices of discrimination of long-Duration, phoneticallysimilar vowels in children with typical and atypical language development. J.Speech Lang. Hear. Res. 53 (3), 757–777.

Delorme, A., Makeig, S., 2004. EEGLAB: an open source toolbox for analysis ofsingle-trial EEG dynamics including independent component analysis. J.Neurosci. Methods 134 (1), 9–21, http://dx.doi.org/10.1016/j.jneumeth.2003.10.009.

Espy, K.A., Molfese, D.L., Molfese, V.J., Modglin, A., 2004. Development of auditoryevent-related potentials in young children and relations to word-level readingabilities at age 8 years. Ann. Dyslexia. 54 (1), 9–38.

Hämäläinen, J.A., Guttorm, T.K., Richardson, U., Alku, P., Lyytinen, H., Leppänen,P.H.T., 2013. Auditory event-related potentials measured in kindergartenpredict later reading problems at school age. Dev. Neuropsychol. 38 (8),550–566.

Häyrinen, T., Serenius-Sirve, S., Korkman, M., 1999. Lukilasse. PsykologienKustannus Oy, Helsinki, Finland.

Kaplan, E., Goodglass, H., Weintraub, S., Segal, O., Laine, M., Koivuselkä-Sallinen, P.,Niemi, J., 1997. Bostonin Nimentätesti. Waverly, Inc./Psykologien KustannusOy., Helsinki (Boston Naming Test).

Karhu, J., Herrgård, E., Pääkkönen, A., Luoma, L., Airaksinen, E., Partanen, J., 1997.Dual cerebral processing of elementary auditory input in children. Neuroreport8 (6), 1327–1330.

Kirby, J.R., Georgiou, G.K., Martinussen, R., Parrila, R., 2010. Naming speed andreading: from prediction to instruction. Read. Res. Q. 45 (3), 341–362, http://dx.doi.org/10.1598/RRQ.45.3.4.

Korkman, M., Kirk, U., Kemp, S.L., 1997. NEPSY: Lasten NeuropsykologinenTutkimus. Psykologien Kustannus Oy, Helsinki (NEPSY: A developmentalneuropsychological assessment).

Korkman, M., Kirk, U., Kemp, S.L., 2008. Nepsy II—Lasten NeuropsykologinenTutkimus. Psykologien Kustannus Oy, Helsinki (NEPSY II—A developmentalneuropsychological assessment).

Kuhl, P.K., Conboy, B.T., Coffey-Corina, S., Padden, D., Rivera-Gaxiola, M., Nelson, T.,2008. Phonetic learning as a pathway to language: new data and nativelanguage magnet theory expanded (NLM-e). Philos. Trans. R. Soc. London B:Biol. Sci. 363 (1493), 979–1000.

Kuhl, P.K., 2004. Early language acquisition: cracking the speech code. Nat. Rev.Neurosci. 5 (11), 831–843.

Kushnerenko, E., Ceponiene, R., Balan, P., Fellman, V., Huotilaine, M., Näätäne, R.,2002. Maturation of the auditory event-related potentials during the first yearof life. Neuroreport 13 (1), 47–51.

Kuuluvainen, S., Nevalainen, P., Sorokin, A., Mittag, M., Partanen, E., Putkinen, V.,. . . Kujala, T., 2014. The neural basis of sublexical speech and corresponding

nonspeech processing: a combined EEG–MEG study. Brain & Language 130,19–32.

Kuuluvainen, S., Alku, P., Makkonen, T., Lipsanen, J., Kujala, T., 2015. Cortical speechand nonspeech discrimination in relation to cognitive measures in preschoolchildren. Eur. J. Neurosci., http://dx.doi.org/10.1111/ejn.13141, n/a-n/a.

2 Cogni

L

M

(

P

P

R

S

S

S

32 S. Kuuluvainen et al. / Developmental

ovio, R., Näätänen, R., Kujala, T., 2010. Abnormal pattern of cortical speech featurediscrimination in 6-year-old children at risk for dyslexia. Brain Res. 1335,53–62.

cGettigan, C., Scott, S.K., 2012. Cortical asymmetries in speech perception: what’swrong, what’s right, and what’s left? Trends Cogn. Sci. 16 (5), 269–276, http://dx.doi.org/10.1016/j.tics.2012.04.006.

OSF) Official Statistics of Finland, 2013. Educational structure of population.Statistics Finland (e-publication).

ihko, E., Kujala, T., Mickos, A., Antell, H., Alku, P., Byring, R., Korkman, M., 2005.Magnetic fields evoked by speech sounds in preschool children. Clin.Neurophysiol. 116 (1), 112–119.

onton, C.W., Eggermont, J.J., Kwong, B., Don, M., 2000. Maturation of humancentral auditory system activity: evidence from multi-channel evokedpotentials. Clin. Neurophysiol. 111 (2), 220–236.

abiner, L.R., Schafer, R.W., 1978. Digital Processing of Speech Signals, vol. 100.Prentice Hall, Englewood Cliffs.

hafer, V.L., Morr, M.L., Datta, H., Kurtzberg, D., Schwartz, R.G., 2005.Neurophysiological indexes of speech processing deficits in children withspecific language impairment. J. Cogn. Neurosci. 17 (7), 1168–1180.

hafer, V.L., Ponton, C., Datta, H., Morr, M.L., Schwartz, R.G., 2007.Neurophysiological indices of attention to speech in children with specificlanguage impairment. Clin. Neurophysiol.: Off. J. Int. Fed. Clin. Neurophysiol.

118 (6), 1230–1243, http://dx.doi.org/10.1016/j.clinph.2007.02.023.

hafer, V.L., Yu, Y.H., Wagner, M., 2015. Maturation of cortical auditory evokedpotentials (CAEPs) to speech recorded from frontocentral and temporal sites:three months to eight years of age. Int. J. Psychophysiol. 95 (2), 77–93, http://dx.doi.org/10.1016/j.ijpsycho.2014.08.1390.

tive Neuroscience 19 (2016) 223–232

Sorokin, A., Alku, P., Kujala, T., 2010. Change and novelty detection in speech andnon-speech sound streams. Brain Res. 1327, 77–90.

Sussman, E., Steinschneider, M., Gumenyuk, V., Grushko, J., Lawson, K., 2008. Thematuration of human evoked brain potentials to sounds presented at differentstimulus rates. Hear. Res. 236 (1-2), 61–79, http://dx.doi.org/10.1016/j.heares.2007.12.001.

Tervaniemi, M., Hugdahl, K., 2003. Lateralization of auditory-cortex functions.brain research. Brain Res. Rev. 43 (3), 231–246.

Torppa, M., Lyytinen, P., Erskine, J., Eklund, K., Lyytinen, H., 2010. Languagedevelopment, literacy skills, and predictive connections to reading in Finnishchildren with and without familial risk for dyslexia. J. Learn. Disabil. 43 (4),308–321, http://dx.doi.org/10.1177/0022219410369096.

Tremblay, K., Kraus, N., McGee, T., Ponton, C., Otis, B., 2001. Central auditoryplasticity: changes in the N1-P2 complex after speech-Sound training. EarHear. 22 (2), 79–90.

Vouloumanos, A., Werker, J.F., 2004. Tuned to the signal: the privileged status ofspeech for young infants. Dev. Sci. 7 (3), 270–276.

Vouloumanos, A., Werker, J.F., 2007. Listening to language at birth: evidence for abias for speech in neonates. Dev. Sci. 10 (2), 159–164.

Wechsler, D., 1999. WISC-III—Wechsler Intelligence Scale for Children—III. ThePsychological Corporation, San Antonio, TX, U.S.A.: Psykologien kustannus,Helsinki, Finland.

Wechsler, D., 2010 WISC-IV—Wechsler Intelligence Scale for Children— IV. NCSPearson Ltd., London, U.K.: Psykologien kustannus, Helsinki, Finland.

Wecshler, D., 2009. WPPSI-III—Wechsler Preschool and Primary Scale ofIntelligence —Third Edition. The Psychological Corporation, San Antonio, TX,U.S.A.: Psykologien kustannus, Helsinki, Finland.